Using the institution as the unit of analysis, this chapter examines persistence rates for hundreds of institutions that have administered National Survey of Student Engagement (NSSE). A model predicting persistence with institutional and aggregated student characteristics was used to determine two groups of institutions: those that are doing better than expected and those doing as expected in terms of undergraduate student persistence rates. Student engagement was compared between the two groups to determine the links between doing better than expected and effective educational practices, particularly practices connected to the classroom. The authors also compared faculty practices between the better-than-expected and as-expected institutions by limiting their sample of institutions further to those that have administered the Faculty Survey of Student Engagement (FSSE). The results of this study suggest an institution-level analogue to the assertion that a student's development is a function of an optimal blend of challenge and support. Institutions with better-than-expected persistence had higher levels of academic challenge and were viewed by students as more supportive, on average, compared to institutions with as-expected persistence rates. The implication is to simultaneously ratchet up the level of academic challenge as well as the amount of support provided to students. (Contains 2 tables.)